Add functionality to load/save distisets to/from disk

[plaguss]

Description

This PR includes two new methods on Distiset to save the content to disk (or load from it), and also to a remote storage based on the content of the storage_options passed, following the implementation of Hugging Face's datasets: https://huggingface.co/docs/datasets/filesystems#saving-serialized-datasets

with Pipeline(...) as pipe:
    distiset = pipe.run()

distiset.save_to_disk(distiset_path)  # Or include the storage_options for the remote storage.
ds = Distiset.load_from_disk(distiset_path)

[plaguss]

@rasdani did you have time to try saving/loading to s3 with this branch?

[rasdani]

yes, saving and loading with S3 works so far! :)

However one needs to keep in mind, to append default/train (or your respective config/split name) to your S3 directory path.
This is expected behaviour from datasets.load_from_disc() though.

EDIT: hold on, I just realised, I tested with datasets.load_from_disc() and not Distiset.load_from_disc(). Will try in a minute.

[gabrielmbmb]

LGTM!